Tag: Waveform

Weirdly, I haven’t written much about synchrophasors on this blog. This is unfortunate, because there are some things we need to talk about. This topical gap hasn’t been for lack of desire or lack of material. Transparently, I had such a backlog of synchrophasor content that the thought of slogging through all of it just to provide background for some potential future essay couldn’t compete with the more broadly relatable material I also wanted to write about. Through my procrastination, however well meaning, we’ve now found ourselves in a position where I don’t have reference material to build on for this essay. So, here we are. You’re welcome. I think…

Either way, we still need to talk about synchrophasors. I’m going to have to assume the reader has some substantial technical and historical familiarity with the topics herein. If that leaves you with questions, fair enough. Now I have an incentive to come back and fill in the gaps.

Disclaimers done. So, what is it that we need to talk about?

Stop Sh*tting on Synchrophasors

Over the last few years, I’ve noticed a frustrating trend in the narratives emerging from the waveform community. Compoundingly, my employees come back from conferences reporting versions of the same thing. This is what we hear: In making the case for waveform data, some academics, researchers, and vendors seem determined to tell everyone that synchrophasors are inadequate. Bad. Garbage. Something we should stop spending our time on because the better thing has arrived.

Now, I’m not quoting any one person or organization individually. But my general characterization of the message is fair. Moreover, I’ve heard enough versions of it to recognize the pattern.

My message back is this: STOP SH*TTING ON SYNCHROPHASORS!

OK. Breathe…

There are two primary issues with the synchrophasor-critical narrative.

First and foremost, some of the pro-waveform/anti-PMU rhetoric is based on some scientifically and mathematically incomplete ideas. The technical argument behind this is outside the scope of this essay, but a close friend and colleague, Dr. Chetan Mishra, has published his preliminary work addressing it here: arXiv.org and here: ResearchGate. The cliff notes version is that there are many claims about the supposed invisibility of fast dynamics due to some intrinsic attenuation but it turns out that in the real-world, there is a lot of desirable information still observable post-attenuation. I won’t butcher his thesis more than that right now.

The second issue, and the subject of this essay, is strategic. The rhetoric is doing a tremendous disservice to everyone in the industry. It is, of course, undermining the synchrophasor mission. But ultimately, whether you see it or not, you are also undermining your own mission to bring waveform and ultra-high resolution sensing capabilities to bear on the industry. I said it clearly: the synchrophasor-critical narrative being used as advocacy for waveform data will ultimately undermine and delay the adoption of the very technology it seeks to advance.

Don’t get me wrong. I’m not only pro-synchrophasor. I’m pro-waveform. Heck, I am pro-grid sensor data. That is precisely why this bothers me. But the synchrophasor mission is not over. We have not institutionalized a synchrophasor future for everyone. Making those capabilities sound obsolete or inferior before we have finished bringing them to bear is no way to help the industry progress. And if we take our eye off the ball now, our accomplishments thus far will succumb to entropy.

I now recognize this particular mistake because I have committed the same sins myself.

I Was There for the Rhetoric

So much of the synchrophasor narrative centered on demonstrating the shortcomings of SCADA telemetry. SCADA is too slow to show us this or that. It’s not time-synchronized. It’s a scalar measurement. Look at all the phenomena we could see if we had synchrophasors instead. I could go on. Those are some of my favorites. The comparisons clearly demonstrated real technical advantages, specifically from a signals perspective. Somewhere along the way, and perhaps more implicitly than explicitly, we attached an unnecessary condition to our success: SCADA had to lose for synchrophasors to win.

Now a proper synchrophasor historian (a person not a database – pun intended) might tell me that replacing SCADA was never the original intention. They may be right. But I was there for the rhetoric. I contributed to the rhetoric. I don’t need to establish that every person in the community believed in replacement to recognize that we spent a lot of energy making our case by contrasting technical features with what came before.

Synchrophasors never needed to dethrone SCADA. Every new capability should not require the utility to repudiate its previous investment. SCADA does many of the things it was designed to do very well. Asking it to do something else and then declaring it deficient was an easy technical demonstration. It was also an incomplete account of what utilities needed and what adopting another capability would require. We had enough work to do explaining the value, building the infrastructure, changing practices, and making these capabilities sustainable. Turning the existing measurement system into an opponent did not make that work easier.

I now believe that our challenges to SCADA as a core part of our narrative were one reason our transformation has taken as long as it has. I also now believe that an inversion of this thinking will secure our future. I don’t want to watch history repeat itself and have the next generation of measurement technology inherit this same flaw unnecessarily. Please keep reading.

I Like Waveforms, Too

There are good reasons to want waveform data. I want it, too. Lots of it. There are things that higher-resolution measurements can reveal that slower measurements can’t. We should be earnest about those capabilities and where they fit.

The strongest version of the waveform argument is worth taking seriously. The pitch is simple: if the waveform contains what you need to derive a synchrophasor, or any other grid measurement, because fundamentally the conductor signal is itself a waveform, then why not collect the more fundamental measurement? Why commit to a particular representation when you could preserve more of the underlying signal and decide what to derive from it later? Honestly… I agree!

However, there are two essential problems with this pitch:

1. A better representation of a signal is not, by itself, a reason to supersede an existing measurement system.
2. Promising something better around the corner gives utilities a reason to postpone useful work today.

Let’s start with the second issue and work backwards.

The Future Can Wait. Apparently…

Even if we temporarily grant the entire promised waveform future, what happens between now and then? How long will it take to deploy these capabilities at the scale being imagined? Five years? Ten? Twenty? Those are questions, not forecasts. But they are questions that need answers before “the future will be better” becomes a reason to stop working on the present. The fear that this could lead to us discontinuing our current work is a real fear. And it is because of classic utility culture (which sometimes I think is just human culture with transmission lines).

Utilities love to kick the can. They are very good at it, too. Utility engineers and decision-makers can find plenty of reasons to delay difficult changes, especially ones involving capabilities they haven’t yet figured out how to institutionalize. Now, we are handing them another reason: perhaps they shouldn’t do the hard work on synchrophasors because waveform data will solve the problem later. I’m telling you: this kind of idea is like catnip for the utility industry.

My concern is that this is an appealing story precisely because it allows us to wait. Present needs remain unaddressed while we put our faith in a future deployment, future infrastructure, and future organizational readiness. A compelling destination does not, by itself, justify postponing the decisions required today.

If you want waveforms adopted, you should care about this. The habit of waiting for the next capability can outlive whichever capability happens to be next. The synchrophasor journey is the same treadmill that will power the waveform journey.

The Grid Wasn’t Measured in a Day

There is a deeper problem with the replacement argument. We tend to compare SCADA, synchrophasors, and waveforms through the lens of signal processing. It goes like this: begin with a common source signal, compare what each representation preserves, and determine which one gives us the best account of that source. Perhaps non-coincidentally, this feels a lot like a provenance argument, too.

But the measurement landscape of an actual grid did not originate from one common signal, at one location, observed through several interchangeable instruments. The sensors are distributed through an inherited physical and institutional system. What each modality contributes depends partly on the footprint that history produced. The SCADA points across a service territory did not appear with the snap of a finger. Engineering standards were defined and redefined. Equipment was built and replaced. Measurements were installed under the constraints, incentives, and technical capabilities of their time. What we observe today encodes those decisions, not only the underlying signal(s).

The same is true of synchrophasors. In my experience, there was no single project that went through the entire service territory and replaced every SCADA point with a higher-resolution equivalent. We defined engineering standards, and substations got PMUs as we visited them for capital work. The resulting deployment reflected where work happened while a particular standard was in effect. Waveform deployments will encounter their own version of this reality. We will put sensors where we can, when we can, under standards and investment decisions that will themselves continue to change. The measurements may be fast. Deploying the measurement system will take time.

The same observation extends beyond utility-owned sensors. Contemporary Data-as-a-Service or adjacent tech companies give us other digital lenses and other locations from which to observe the grid. Their measurements emerge under different economic constraints and incentives. What gets measured, where it gets measured, and what remains available are all influenced by the context of the deploying institution.

What and how we measure is therefore a function of much more than which algorithm best reconstructs some fundamental grid signal. It also reflects when the installation opportunity appeared, what the engineering standard required, what the investment climate supported, and what was already there.

Centuries in a Glass. Decades in the Database.

In my previous post, I wrote how on a recent trip to France I had been expecting champagne to simply be an exclusive label but ended up encountering something that felt remarkably familiar: engineering standards. A system for making something particular, reliably, while preserving room for natural, meaningful variation. The method mattered. So did the place, the season, the intentions carried forward, and the time in which things had to happen. The standards could establish a bar and give a name to the result, but they operated within a landscape. Slope, soil, moisture, temperature, and the conditions of a particular year all participated in what eventually arrived in the glass. Learning about that formation became part of experiencing the product.

Our measurement landscape has its own conditions of formation. Engineering standards establish what we intend to install. Capital work creates opportunities. Geography determines where those opportunities occur. Economics, institutional boundaries, and changing priorities influence what gets measured and what gets preserved. Different generations of decisions accumulate across time and space until they arrive in the record we have available now. Consider what we mean when we say that one measurement lets us “see more”. Higher temporal resolution can show us more of what happened within a particular interval. But where are those measurements? How far back do they go? How long do we retain them? What quantities were measured in the first place?

Relative costs influence deployment and retention. We may have a much faster measurement at fewer locations, with a shorter retained history. A slower measurement may cover more of the territory and reach much farther into the past. These are different dimensions of observation. Bringing them together gives us context that a comparison of sampling rates doesn’t capture. We also have to look beyond the measurements arriving right now to the accumulated historical record. Slower data contains observations from periods before the faster capabilities were installed. No matter how impressive the new sensor is, deploying it today cannot retroactively collect the past. Nor can it be deployed arbitrarily in the here and now.

On my champagne adventure, I found myself contemplating centuries in a glass. With grid data, we’ve got decades in a database. The method of its formation is also encoded in what we can extract from the historical record. We should be very careful about dismissing part of that measurement inheritance because another part is faster.

With A.I., No Sensor Left Behind

Imagine that we could accommodate all generations of measurement together, including those yet to be deployed. Would we then voluntarily discard useful observations merely because they came from the older, slower paradigm? Would we insist that everything conform to whichever measurement standard happens to be fashionable today? Of course, particular sensors can become redundant. Particular systems can deserve retirement. But why should age or relative speed of a modality disqualify the observations it contributes before we have considered their relationship to the rest of the record?

This is where A.I. makes the picture much more interesting. With A.I., we can begin working with the full breadth of that history, including information encoded in how the measurement system came into being. Where we measured. What we measured. When a location became observable and what was happening elsewhere before it did. The accumulated decisions about standards, deployment, and retention have shaped the record.

I don’t mean that A.I. can magically reconstruct measurements we never collected. I mean that its possibilities invite us to consider the historical record as a whole, across modalities, locations, and generations of measurement systems. We can ask questions whose answers may depend on relationships we haven’t yet thought to examine. “Decades in a database” starts to look like more than a collection of old measurements. What’s slow is new again because we have new possibilities for exploring what that history contains.

But contemplating the possibilities is the easy part. We need to be able to experiment with them. One of the things I have come to know about succeeding with synchrophasors is that the cost of experimentation has to be low. If every question requires an expensive exercise in finding, extracting, moving, and preparing data before anyone can test an idea, we will explore fewer ideas, or none at all. The right data infrastructure changes that equation favorably. It makes trying something, and finding out whether it is useful, a much more achievable undertaking. The infrastructure also matters because “consider decades in the database together” is otherwise just an aspiration. We need somewhere for these different forms of data to coexist, retain their meaning, and become available for analysis across their different timescales. An application that sees one modality in isolation cannot supply that relationship merely by having a better algorithm.

The Inversion: What’s Slow is New Again

We should want those capabilities for slower data, too. This is where my thinking has inverted. Everyone has slower data. Fewer people have higher-resolution data, and they are at different points in figuring out how to use it. Why should access to the experimentation capabilities we want for synchrophasors depend on having already made that journey? Bring slower data into the high-speed platform as a primary value proposition. Make it useful for the measurements people already have and the questions they need to answer today. Then, when they are ready for higher resolution data, they have somewhere to put it and a way to begin working with it. We can serve an immediate need while creating an on-ramp to the faster capabilities.

I began my synchrophasor journey helping make the case that we needed something faster than SCADA. I still believe in the value of those faster measurements. But I now believe that fulfilling the synchrophasor mission requires earnestly embracing the slower data I once helped deride. A.I. pushes us to consider the full, multimodal history as a baseline capability. The right infrastructure minimizes the cost of experimenting with that history, starting with the data people already have.

That is the inversion. Serving the slower measurement world helps establish the infrastructure and practices through which faster capabilities can become useful. The same thinking should matter to waveform advocates: making the synchrophasor world work better is part of getting more waveforms adopted sooner.

So, please stop sh*tting on synchrophasors. Have a glass of champagne and take a look at the big picture. In an A.I. future, what’s slow is new again.

Which leaves a practical question: what are we actually going to do to embrace slow data?